AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

Distilling Knowledge from Large-Scale Image Models for Object Detection

Gang Li, Wenhai Wang, Xiang Li, Ziheng Li, Jian Yang, Jifeng Dai, Yu Qiao, Shanshan Zhang

ECCV 2024 · Label-efficient and robust learning › Knowledge distillation · approach: Knowledge distillation for detectors · 7 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use a frozen-backbone teacher instead of a fully fine-tuned teacher with Query Distillation (ablation)DLIM-Det (Swin-T student) · COCO val2017mAP: 51.6 -> 52.2Improved
inherit teacher detection head parameters (ablation)DLIM-Det (Swin-T student) · COCO val2017mAP: 54.4 vs 52.2 without head inheritingImproved
freeze teacher backbone while matching teacher mAP (56.7 vs 56.9) (ablation)Swin-L teacher · COCO val2017student mAP: 53.1 -> 54.4Improved
freeze backbone in teacher trainingSwin-L teacher detection performance · COCO val2017teacher mAP: 58.1 -> 56.6Hurt
add Query Position Distill and Query Relation Distill together (ablation)Swin-T student with Swin-B teacher · COCO val2017mAP: 51.3 -> 52.1Improved
apply Query Distillation on ResNet-50 studentDINO ResNet-50 · COCO val2017mAP: 48.9 -> 49.8Improved
distill on all 6 decoder layers instead of fewer layers (ablation)Query Distillation · COCO val2017, Intern-T studentmAP: 54.1 -> 54.5Improved
distill from randomly sampled 300 teacher queries instead of all 900 (ablation)Query Distillation · COCO val2017, Intern-T studentmAP: 54.2 -> 54.5Improved
distill from foreground-only or background-only teacher queries instead of random sample (ablation)Query Distillation · COCO val2017, Intern-T studentmAP: 54.5 -> 54.2Hurt
use larger teacher (Swin-L frozen instead of Swin-B frozen) (ablation)DLIM-Det · COCO val2017, Swin-T studentmAP: 53.5 -> 54.4Improved
apply prior KD (FGD) with InternImage large teacherInternImage-T student · COCO val2017mAP: 53.2 -> 52.8Hurt
distill directly from large teacher instead of two-step teacher assistantSwin-T student · COCO val2017mAP: 53.7 -> 54.4Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy